Nvidia Rejects Calls to Slow AI Development Race
Jensen Huang pushed back against a proposed AI slowdown, exposing a widening split between infrastructure sellers and safety-focused model builders.
Nvidia CEO Jensen Huang rejected calls for a broad slowdown in AI development during an appearance at the All-In Summit on September 14, saying the company would not allow such a pause to derail the industry’s momentum. President Donald Trump joined the conversation by phone and also dismissed the idea that AI development should be halted because of existential-risk concerns.
A visible industry divide
The remarks followed a weekend in which Anthropic CEO Dario Amodei argued that frontier-model progress should be paced more carefully. OpenAI CEO Sam Altman and other industry figures also acknowledged the need for stronger safeguards. Huang’s position was materially different: he characterized extreme AI-doom claims as unsupported and emphasized the economic and national-security benefits of continued deployment.
The disagreement is not simply philosophical. Nvidia supplies the accelerators and networking systems that make large-scale model training and inference possible. Its commercial interests are therefore tied to continued spending by cloud providers, labs and governments. A slower development cycle could reduce near-term demand for the infrastructure ecosystem, even if it improved safety evaluation.
Why it matters
The exchange makes the policy conflict unusually explicit. Model developers are warning that capability gains may be arriving faster than testing and governance systems can absorb, while the leading hardware supplier is arguing that slowing progress could weaken competitiveness, particularly against China. The result is a debate over whether safety should be achieved through a coordinated pause, incremental controls or continued deployment with stronger safeguards.
No government-wide slowdown was announced, and Huang’s comments do not change Nvidia’s product roadmap. But the public disagreement matters because it shows that “AI safety” is no longer a unified industry position. The companies building models, selling infrastructure and setting policy now have visibly different incentives.
Uncle Cat take
Huang’s “no slowdown” stance carries unusual weight because Nvidia benefits from every additional training cluster. That does not make his warning wrong, but it makes his preferred timetable impossible to treat as neutral evidence.